Bulletin, June/July 2006


Toward an Enriched (and Revitalized) Sense of Help: Summary of an ASIS&T 2005 Panel Session

 

by Stephanie W. Haas, Laurie Brown, Sheila Denn, David Locke and Ben Shneiderman

Stephanie W. Haas, University of North Carolina, stephani<at>ils.edu.com

Laurie Brown, Social Security Administration, Office of Policy, laurie.brown<at>ssa.gov

Sheila O. Denn, University of North Carolina, denn<at>ils.unc.edu

David Locke, Wordsmith LLC, david<at>wordsmith.net

Ben Shneiderman, University of Maryland, College Park, ben<at>cs.umd.edu>

In January 2005, the NSF-sponsored GovStat project (http://ils.unc.edu/govstat, Gary Marchionini, PI) sponsored a symposium on help. The goal of the symposium was to discuss what project members, federal agency colleagues and others had learned about the problems with existing help facilities for large public-access websites and to start forming a new vision of what help could and should be. We agreed that we must rejuvenate interest in help and develop a research agenda to address the gaps in our understanding. At the 2005 ASIS&T meeting in Charlotte , North Carolina , we reunited symposium participants to share and discuss the symposium findings with the ASIS&T community.

We opened the panel session by asking the audience who among them had clicked on a “help” button or link during the last week. Of the 30 or 40 who raised their hands, only two responded that they had found what they needed. This set the stage for the panelists’ perspectives: what is the problem, and what can the ASIS&T community do to solve it?

The Panelists’ Perspectives

            Panelists were asked to talk about challenges involved in providing good help, as well as success stories and opportunities for the future.

David Locke

David Locke, a consultant in online information systems at WordSmith, LLC, who was unfortunately unable to attend the conference, contributed a framework for designing help that includes three critical components: users; information types; and signaling, affordances and access. Stephanie Haas presented an overview of his ideas on “Functional Help Design Criteria.”

In Locke’s view help is a collection of answers, either to questions users have asked in the past or to questions we anticipate they might have now or in the future. Before we can prepare the answers, we need to know who our users are and what brings them to help. We already know that users do not come readily or happily to help, and quick access to the information they seek is required to make help functional. Fundamental divisions of the answer collection may start with the distinction between definitions and procedures that relate to the software interface and those that relate to the content the interface presents.

What if we view an answer as a topic – information that is relevant to the user’s question (as well as similar questions)? For the content provider, an answer is then a collection of discrete units of information that have been gathered and organized to address this particular topic. Different topics, or answers, are created when units are joined in different combinations. As defined by the IEEE Learning Technology Standards Committee in Learning Object Metadata (LOM) Final Draft Standard IEEE 1484.12.1-2002, learning objects are a model for these information units. Learning objects are defined there as “any entity, digital or non-digital, which can be used, re-used or referenced during technology supported learning.”

In addition to the answers or topics, we also need to develop clearer signaling methods that will help users see what kind of information is available and how to get to it quickly. Absent these signaling conventions, users spend frustrating first-contact moments trying to figure out how to use help and what they will find before they can discover the help content itself. Often, current help systems seem to layer one potentially frustrating information system on top of an already frustrating one.

This framework suggests several strands of research.

·        What are cost-efficient ways of learning about users and their questions? Will the same methods work for the universe of Web users and a homogeneous set of users in a constrained environment?

·        Going beyond the interface/content division, should organizational structures to support browsing and searching for help topics differ from structures for other kinds of content?

·        What do we gain or lose by viewing the design of help as a problem of defining the basic information unit structure, identifying units that are relevant to particular questions and combining them into coherent topics?

·        How can we accelerate establishment of conventions for signaling access and navigation of help?

Ben Shneiderman

Ben Shneiderman, professor of computer science and founding director (1983-2000) of the Human-Computer Interaction Lab at the University of Maryland , described two innovative approaches to providing procedural, interface help: Show me! and multi-layered interfaces. These approaches flesh out aspects of the research framework. ShowMe! shifts the user attitude from the desperate sense of help to the more user-centered notion of requesting information. ShowMe!s are short (one-minute) narrated demonstrations of the interface in action while carrying out meaningful tasks, combined with an audio narration that describes what the users see. Users can stop and replay repeatedly. ShowMe!s are created with screen capture tools such as Camtasia, thereby providing higher resolution and smaller downloadable files than video (more info at: www.cs.umd.edu/local-cgi-bin/hcil/sr.pl?number=HCIL-2005-02).

Multi-layered interface designs enable first-time and novice users to begin with a limited set of features at the first layer. They can remain at the first layer, then move up to higher layers when needed or when they have time to learn further features. While there are interesting design problems in how to define the layers (consider, for example, layering by function or by application or task), this is a very promising strategy that needs further exploration and testing. Such layered designs have been enormously successful in video games, and early examples suggest that they can have high payoffs for advanced desktop, Web and mobile device applications (more at: www.cs.umd.edu/local-cgi-bin/hcil/sr.pl?number=HCIL-2003-33).

These approaches illustrate two basic principles in designing help. First, the user must have control, ranging from starting, stopping and replaying to choosing the specific topic of the help presentation. Second, procedural help can be presented in small, discrete units that highlight specific topics or functions. The user must be able to focus on important features or ideas: an undifferentiated dump of information is overwhelming at best and an intimidating waste of time at worst.

Sheila Denn

Sheila Denn, a doctoral candidate at the School of Information and Library Science, UNC-CH, further explored the distinction between interface help and content help.

A brief history of research into help reveals that most research thus far has been in the area of interface help, defined as support for the use of the features included in a particular piece of software. We can draw a distinction between this kind of help and content help, defined as support for the use of the content and concepts contained within an information system. This distinction is important now because of the explosion of content on the Web, with which users interact using a fairly constrained set of interface features. This necessarily broadens the scope of what help should be to include features of the content as well as the interface. While focusing on exploring the idea of content help and setting out a proposed research agenda in content help, it must be acknowledged that the distinction is a somewhat artificial one, and that in the end interface help and content help are necessarily intertwined.

Content help can include support for general concepts within the domain of the information system as well as support for particular instances of those concepts. This support can include definitions of concepts, guidelines for usage and manipulation of concepts, and so forth. Most examples of good content help in the current environment occur outside the online information system and are human-mediated, such as interactions with reference librarians or with help lines provided by the manufacturers of consumer goods, such as home appliances. Perhaps the best example of good online content help is that provided by income tax software such as TurboTax (www.turbotax.com), where context-sensitive help is provided for individual items as users proceed through preparing their tax forms. Note that this is a very tightly constrained domain with a very highly specified task, where the users’ needs, goals and potential problems are relatively homogeneous and predictable. One open question is whether effective content help can scale to more broadly conceived information systems where user tasks cannot be predicted.

One concept related to help that should inform our research is that of usability. It is widely recognized that providing support for the use of interface features within a software application is essential to rendering that software usable. So the question becomes: How do we provide support for information use as an essential piece of information usability? What does it mean for information to be usable?

A research agenda for content help must guide us to understand the complex interactions between user and task characteristics, between content and interface characteristics, and amongst all four of these kinds of characteristics to make up information use. One of the first challenges we must face is in research design. How do we design studies to explore these interactions? What kinds of metrics can we use to measure them? Once we have designed and undertaken such studies, how can the results be applied to the design of integrated interface and content help? There is a need for the human-computer interaction and information use communities to come together to resolve some of the methodological issues and move the field of help forward.

Laurie Brown

Laurie Brown, webmaster for the Social Security Administration's (SSA's) Office of Policy discussed the challenges and opportunities presented by the Web to federal statistical agencies and the potential of rich Internet applications in providing better online help to users.

SSA's Office of Policy is responsible for conducting policy research and, as home to one of the principal federal statistical agencies (the Office of Research, Evaluation and Statistics or ORES), is also responsible for the dissemination of data and statistics about the programs SSA administers. That latter task includes publishing a series of statistical compilations that include about 1000 data tables per year. The Office of Policy has a well-established print publication program, as do most federal statistical agencies. However, the advent of the Web presented new opportunities and challenges.

The Web has allowed the federal statistical agencies to reach a wider audience in ways they weren't able to before. While such unfettered access to information is of benefit to many, it also increases the possibility of misinterpretation by those not trained in statistics or familiar with specific federal agencies or programs. In addition, the nonsequential nature of the Web means that the traditional structure of statistical publications – perhaps a brief introduction at the front, a glossary at the back and lots of tables – isn't appropriate anymore. Better ways are needed to help users find what they are looking for and to understand what they find.

In order to create an enriched Web experience for users, help models must be backed by an institutional commitment to improving the user experience and the technology to do so. The development of stovepipes within and between agencies has led to several issues that are highlighted when information is collectively presented on the Web:

  • Data harmony – different numbers being reported for the same concept.
  • Definitions – the same term being used for different concepts.
  • Labeling – different terms being used for the same concept.

Before effective help can be provided for users, agencies must harmonize their data – identifying a single, authoritative source for each concept – and fully develop accompanying metadata.

The current HTML page-based model used on the Web also presents challenges when trying to provide help to users. That model is not conducive to helping users stay focused on their task while seeking more information or providing that information in context. Constant server calls and page refreshes distract users from their tasks. In addition, the common ways of delivering help each pose challenges for users:

  • New page – users completely lose the context of their task.
  • New window – focus shifts from user's task and new window may obscure relevant parts of original window.
  • Frames – take up valuable screen real estate whether or not help is needed at that moment.

Also, HTML's limited set of attributes sometimes puts accessibility and usability at odds. For example, the "alt" attribute of an image tag may be used to provide equivalent text for a blind user's screen reader or a tool tip for a sighted user's graphical browsers, but it can't do both.

Emerging technologies, such as rich Internet applications (RIAs), address many of the limitations associated with static HTML and offer a lot of potential in providing better help to users. RIAs provide richer user interfaces, allow greater interaction with applications and provide immediate feedback to users without the need for constant calls back to the server and frequent page refreshes. Components within an RIA allow for the progressive disclosure of information and smooth transitions between application states so users don't lose the context or focus of their task. The use of XML with RIAs would allow for the customization of help; instead of static glossary definitions, users could be presented with more customized information.

The Audience Joins In

            It was clear from the questions and comments made by those attending the session that 1) there is a lot of pent-up frustration about the current state of help, 2) examples of good help exist and 3) it is high time that the various interested parties (researchers, developers, content providers and others) focus on improving help. Participants related their own experiences, as well as those involving family, friends and co-workers. One thing seems clear: everyone needs help with something at some time, regardless of their level of expertise. Understanding users and their needs is thus a crucial part of the research agenda.

Identify and remove barriers. The name help is itself problematic, with connotations of insufficiency and failure rather than learning. Users need sufficient motivation to seek help – the need to complete a task or learn information and the belief that they can find what they need without wasting a lot of time and effort. There was agreement that we need to provide good cues or signals as to what help is available. Cues need to be appropriately prominent and labeled, and as mentioned above, establishing conventions could be useful.

Procedure vs. content. Audience members seemed to agree with the idea that we need different approaches for procedural (task) help and content help. In the abstract, the distinction seems intuitive; in practice, it will not be as clear.

We can view tasks as ranging from the highly structured (with few options or choices needed) to the more creative and free-form. At the structured end, good design can obviate the need for much help. A richly scaffolded environment may essentially eliminate the need for help as a separate feature by making the information the user needs to use and understand the interface and content an integral part of the system – a world where no “help” is needed. At the other end, good design is still essential to reduce the need for help, but guidance (support) in making choices and understanding options is still needed. Layering, Show-me! and other means of providing suggestions and examples are useful here.

Content help is harder. Just-in-time, just-enough help provides enough background or explanatory information for the user to complete the task at hand. Long-term learning is not the immediate goal, although it may be a happy side effect. In other situations, a user may want to put in a greater investment of time and effort for long-term learning, in which case the initial task may just be a point of departure. Without the structure of a predictable task (and therefore more predictable questions or information needs), we must look to other sources of guidance.

Providers of help. Sources for online help include content providers, third-party providers and community providers (lists, blogs, external websites). Content providers are an authoritative source for help on their own content, but providing help may not be a high-priority activity. Can they fold the creation of help resources into the content production process? A cooperative model wherein information units or objects are created by content providers and a third party provides the structure to weave them into answers is attractive, but may not be acceptable to all. Content providers (such as federal statistical agencies) are naturally reluctant to blur the boundaries between the information they provide and stand behind and the related information provided by others. The potentially unknown quality of information contributed by community providers must be balanced against convenient access, shared experiences (frustrations as well as successes) and perhaps a sense that someone is really listening.

Guidance for the future. Panelists and members of the audience cited several examples of websites or tools that are designed well enough to virtually eliminate the need for help, as well as those that incorporate useful and usable help facilities. As mentioned above, the online help provided in TurboTax is well integrated with the user’s task. The new natural gas page at the Energy Information Administration (www.eia.doe.gov/oil_gas/natural_gas/info_glance/natural_gas.html) provides easy access to definitions and source information for their statistical data, although it still requires the user to open a separate page to access this information. Further, the drop-down lists and radio buttons suggest alternative ways of viewing the information that might not have occurred to the user and might even provide better answers to his or her question. Another example is the Genworth Financial Retirement Income Gap Calculator (www.gefinancialassurance.net/calculators/incomeGap.mxml),

which provides definitions of terms and clear transitions from one stage of the user’s task to another.

            From these examples, we can develop guidelines for good practice and an understanding of what approaches are suitable in various information environments. But good examples alone are not sufficient. We also need to seek out relevant learning, usability and design theories for guidance. Scaffolding, distributed cognition, enriched cognitive environments and zone of proximal development provide some guidance.

Conclusion

            The session emphasized that the single word help encompasses a vast array of support tools and information. On one extreme, a well designed tool for a straightforward, well-specified task could eliminate the need for any kind of procedural or content help. Users would be guided from step to step without being given the opportunity to go astray or the need to exit the task environment to learn more. Another perspective might provide small discrete nuggets of information (well-labeled and indexed) from which the user may consult a single brief definition or string several together to create a tutorial. A third could embody active cooperation between the user and the information environment, where every element of the interface offers suggestions and opportunities for learning and exploration.

Our goal is to harness some of the energy, interest and, yes, frustration that was expressed during this session. It is high time we take a new look at help and what it could and should be. ASIS&T members are well-suited to take the lead.

We would like to express our thanks to those who attended the session for their lively discussion.

This work was supported by NSF Grant EIA 0131824.